The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Integration optimization: the practice of optimizing how agents connect and hand off to each other, rather than optimizing each agent's performance in isolation, with the goal of reducing unnecessary friction at integration points while maintaining flexibility
Transition cost: the cognitive or operational cost incurred when one process hands off to another, including time lost in reconfiguration, information degradation, and energy spent on context switching
Subtractive optimization: the discipline of making a process better by removing parts of it rather than adding improvements, where the fastest step is the one that does not exist and the most reliable component is the one that was never built
Addition bias: the systematic human tendency to default to adding elements when asked to improve or change something, even when removing elements would be simpler, cheaper, and more effective, with this bias being stronger under cognitive load and when not explicitly cued
Via negativa: the approach of defining something by what it is not, rather than by what it is, particularly in decision-making where subtractive knowledge is more robust than additive knowledge because we can be more confident about what harms us than what helps us
Seven wastes (muda): the seven categories of waste that can be systematically removed from any production process, including transportation, inventory, motion, waiting, overproduction, over-processing, and defects
Over-processing: a type of waste that occurs when more work or features are added than required to achieve the desired outcome, characterized by activities that feel productive but consume resources without creating value
Optimization sprint: a deliberate, time-boxed session dedicated to improving one specific cognitive agent, with a defined target, a measurable baseline, and a documented outcome
Benchmark: a measurable baseline measurement of system performance taken before any optimization changes are applied, used to determine whether subsequent changes actually improved performance
Optimization log: a structured documentation system that records changes made to a system, the rationale behind each change, the observed results, and the timing of each modification, designed to prevent loss of experimental history and enable systematic learning from optimization efforts
Premature optimization: the practice of optimizing a system or component before measuring and identifying the actual bottleneck, resulting in wasted effort applied to non-constraining elements rather than the true constraint that determines system performance
Continuous optimization: the mindset and practice of making ongoing, small adjustments to cognitive systems and agents based on context changes rather than episodic events or performance degradation, characterized by three properties: ongoing rather than episodic, small rather than sweeping, and responsive to context change rather than just performance degradation
Agent lifecycle: the structural parallel between the creation, deployment, maintenance, evolution, and retirement of cognitive agents and the acquisition, practice, internalization, and letting go of knowledge through learning
Agent creation: a deliberate design act that involves identifying a need, specifying a trigger, defining behavior, setting success criteria, and planning for failure modes rather than a moment of decision or wish
Design: the process of devising courses of action aimed at changing existing situations into preferred ones, involving identifying gaps between what exists and what is desired, then constructing artifacts to close those gaps
Agent deployment: the process of transitioning a designed cognitive agent from the design phase into consistent, automatic operation in daily life, which takes weeks to months and follows a predictable curve of difficulty with specific failure modes and requires deliberate scaffolding, monitoring, and support during the first 30 days to avoid infant mortality
Agent deployment timeline: the measurable duration required for a cognitive agent to transition from conscious, effortful execution to automatic operation, typically ranging from 18 to 254 days with an average of 66 days, characterized by an asymptotic curve where early repetitions produce large gains but later repetitions produce diminishing returns
Agent maintenance schedule: a structured cadence of scheduled reviews and inspections of cognitive agents designed to prevent silent degradation, detect drift, and enable timely adaptation or replacement before failure occurs
Silent degradation: the gradual, undetected deterioration of a cognitive agent's effectiveness that occurs during its stable operating phase, characterized by the absence of visible failure symptoms but progressive loss of alignment with original design or current conditions
Maintenance window: a scheduled, pre-announced period during which cognitive agents are systematically inspected, updated, and adjusted for health, characterized by advance scheduling, stakeholder communication, documented procedures, and defined scope and duration
Agent evolution: the process of modifying an existing cognitive agent by adjusting its triggers, refining its steps, removing accumulated elements, and sharpening its purpose while preserving the agent's core identity and structure
Agent replacement: the process of retiring an existing cognitive agent and building a new one from scratch, ending the old agent's identity while designing something fresh informed by learned lessons but unconstrained by previous design decisions
Agent versioning: the practice of explicitly labeling, preserving, and maintaining changelogs for cognitive agents to enable comparison, rollback, and learning from changes by tracking version history, documenting diffs, and storing previous states
Agent retirement criteria: specific, measurable conditions defined in advance under which a cognitive agent will be retired regardless of emotional attachment or sunk cost considerations, serving as pre-committed decision-making frameworks that eliminate the bias toward agent accumulation by establishing explicit triggers for termination